Sidekiq::ExpectedFailures

If you don't rely on standard sidekiq's retry behavior and you want to track exceptions, that will happen one way, or another - this thing is for you.

Installation

Add this line to your application's Gemfile:

gem 'sidekiq-expected_failures'

And then execute:

$ bundle

Or install it yourself as:

$ gem install sidekiq-expected_failures

Usage

Let's say you do a lot of API requests to some not reliable reliable service. Inside your worker you handle Timeout::Error exception - you delay it's execution, maybe modify parameters somehow, it doesn't really matter, what matter is that you want to log that it happen in a convenient way. Describe that case using ruby:

class ApiCallWorker
  include ::Sidekiq::Worker
  sidekiq_options expected_failures: { Timeout::Error => nil } # handle that exception, but disable notification

  def perform(arguments)
    # do some work
    # ...

    # this service sucks, try again in 10 minutes
    rescue Timeout::Error => e
      Sidekiq::Client.enqueue_in(10.minutes, self.class, arguments)
      raise e # this will be handled by sidekiq-expected_failures middleware

    # ensure block or some other stuff
    # ...
  end

You can pass a hash of exceptions to handle inside sidekiq_options. Each key-value pair may consist of:

exception => nil - notifications disabled
exception => integer - fires exception notify when x-th exception happens (on daily basis)
exception => [integer, integer] - same as above but for each value

sidekiq-expected_failures utilizes sidekiq's ExceptionHandler module - so you might want to set some same limits for your exceptions and use Airbrake (for example) as a notification service to inform you that something bad is probably happing.

This is how web interface looks like:

It logs each failed jobs to to redis list (per day) and keep global counters (per exception class as a single redis hash).

Default expected failures

You can configure defaults for all your workers (overridden completely by specifying expected_failures hash inside sidekiq_options - per worker).

Sidekiq.configure_server do |config|
  config.expected_failures = { ExceptionHandledByDefault => [1000, 2000] } # with notification enabled
end

Usage with sidekiq-failures

Just be sure to load this one after sidekiq-failures, otherwise failed jobs will end up logged twice - and you probably don't want that.

Clearing failures

At the moment you have 3 public methods in Sidekiq::ExpectedFailures module:

clear_counters - clears all counters (as I mentioned - it's stored inside single redis hash, but I doubt anyone would like to log more than 500 different exceptions, right?)
clear_old - clears failed jobs older than today
clear_all - clears all failed jobs

This might change in the future to something more sane.

Contributing

Fork it
Create your feature branch (git checkout -b my-new-feature)
Commit your changes (git commit -am 'Add some feature')
Push to the branch (git push origin my-new-feature)
Create new Pull Request